← Back to Skills Marketplace
booynal

Tts Cleanup Checklist

by booynal · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
57
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install tts-cleanup-checklist
Description
统一的 TTS 文本清洗与可选章节拆分技能。用于将 Markdown/TXT 清洗为可朗读版本,支持批量处理、可配置清洗档位、质检检查与报告输出。只要用户提到“清洗文本用于 TTS”“按清洗规则处理章节”“去参考文献/引用”“做可朗读稿”“批量清洗章节”,就应优先调用本技能。
README (SKILL.md)

tts-cleanup-checklist

用于把 Markdown/TXT 文本清洗为适合语音合成(TTS)朗读的版本。核心目标是:不改变原意、提升可读性、提高朗读自然度。

使用方式

当用户要求“清洗用于朗读”时,Agent 直接按本技能执行,不讨论技能制作过程。

硬约束(默认始终生效)

  • 不压缩内容,不做总结式改写
  • 不新增观点,不删除正文有效信息
  • 对不确定项做“最小改动”

处理模式

  • clean_only:仅清洗(默认)
  • split_only:仅章节拆分(无压缩)
  • both:先章节拆分,再按规则清洗

可配置清洗档位

  • reference_trim: off | tail_only | aggressive(默认 tail_only
  • punctuation_tune: off | mild | strong(默认 mild
  • number_style: keep | mixed | spoken_cn(默认 mixed
  • english_alias_policy: keep | remove_parenthetical | transliterate_common(默认 remove_parenthetical
  • long_sentence_threshold: 整数,默认 120

清洗方向(按优先级)

P0 结构清理(默认必做)

  • 去除引用标记(如 [1][^2]
  • 去除 URL 裸链与尾部参考资料区(按 reference_trim 档位)
  • 去除不参与朗读的目录/导航/分隔符噪声

P1 噪声修复(默认必做)

  • 清理 OCR 噪声(乱码、异常重复符号)
  • 修复异常空格(多空格、错位空格、断裂换行)
  • 规范标点(连串标点收敛,中英文标点统一)

P2 可读性转换(默认开启)

  • 中英混排降噪,减少朗读卡顿
  • 常见缩写转可读表达(按 english_alias_policy
  • 保留必要术语,不做过度改写

P3 数字口语化(默认开启)

  • 日期、时间、百分比、数量统一为自然口语读法(按 number_style
  • 避免机械逐字符读数字

P4 韵律优化(默认开启)

  • 长句按语义断句(按 punctuation_tune
  • 在并列、转折、因果处增加停顿边界
  • 断句只优化可听性,不改变语义逻辑

可选章节拆分规则(长文推荐)

优先级从高到低:

  1. 中文:序言/前言/引言第X章后记/结语/尾声/致谢
  2. 英文:Preface/IntroductionChapter XEpilogue/Acknowledgments
  3. 若标题不完整(如单独行 12 下一行是标题),自动拼接为章节名
  4. 若目录中的章节名与正文冲突,优先正文首次有效出现位置
  5. 若无法可靠识别章节,回退为“按语义分段”,并在报告中标注

Agent 执行步骤

  1. 读取输入文本(md/txt,支持单文件或目录批量)。
  2. 若模式为 split_only/both,先执行章节识别与拆分。
  3. 依次执行 P0→P4 清洗(按配置档位)。
  4. 生成输出文件。
  5. 生成“清洗说明/批量汇总报告”。
  6. 执行质检并在报告中记录结果。

质检清单(默认必做)

  • 是否残留参考文献段(关键词/编号检测)
  • 是否存在异常长句(超过 long_sentence_threshold
  • 章节连续性(是否漏章/重复章,仅拆分模式)
  • 每文件字符数统计(不含空白)

单文件输出格式(固定)

# 清洗后文本
(正文)

---

# 清洗说明
- 输入文件:\x3Cpath>
- 模式:\x3Cclean_only|split_only|both>
- 应用清洗项:
  - P0: ...
  - P1: ...
  - P2: ...
  - P3: ...
  - P4: ...
- 参数:
  - reference_trim: ...
  - punctuation_tune: ...
  - number_style: ...
  - english_alias_policy: ...
  - long_sentence_threshold: ...
- 质检结果:
  - 参考文献残留:\x3C通过|告警>
  - 异常长句:\x3C数量>
  - 字符数(去空白):\x3Ccount>
- 备注:\x3C如有未处理项或不确定项>

批量输出要求

  • 按文件名顺序处理
  • 每个文件产出对应清洗稿
  • 追加汇总报告,至少包含:成功数/失败数/跳过数、失败原因分类、清洗项命中统计、质检汇总

给 Agent 的提示词模板

请按 tts-cleanup-checklist 规则清洗以下文本用于 TTS 朗读:

输入文件:\x3Cinput_path>
输出文件:\x3Coutput_path>
模式:\x3Cclean_only|split_only|both>
参数:
- reference_trim=\x3Ctail_only>
- punctuation_tune=\x3Cmild>
- number_style=\x3Cmixed>
- english_alias_policy=\x3Cremove_parenthetical>
- long_sentence_threshold=\x3C120>

要求:
1) 按 P0→P4 顺序执行清洗。
2) 不改变原意,不压缩内容,不新增观点。
3) 输出“清洗后文本”+“清洗说明”。
4) 若存在不确定处理,采用最小改动并在备注中说明。
5) 输出质检结果。

批量任务提示词模板

请按 tts-cleanup-checklist 规则批量清洗目录中的章节文件用于 TTS:

输入目录:\x3Cinput_dir>
匹配:\x3Cglob>
输出目录:\x3Coutput_dir>
模式:both
参数:
- reference_trim=tail_only
- punctuation_tune=mild
- number_style=mixed
- english_alias_policy=remove_parenthetical
- long_sentence_threshold=120

要求:
1) 按文件名顺序处理。
2) 每个文件执行 P0→P4 清洗。
3) 每个文件输出对应清洗稿,并生成汇总报告(成功/失败/跳过、失败原因、清洗项统计、质检统计)。
4) 不改写原意,不压缩正文。
Usage Guidance
This skill appears safe for its intended purpose. Before using it, provide only the specific files or directories you want cleaned, set an output directory, and review the generated report to confirm the text was not changed in an unintended way.
Capability Analysis
Type: OpenClaw Skill Name: tts-cleanup-checklist Version: 1.0.0 The skill bundle defines a set of text-processing instructions for an AI agent to prepare Markdown or TXT files for Text-to-Speech (TTS) synthesis. The logic in SKILL.md focuses on removing references, fixing OCR noise, normalizing punctuation, and splitting chapters without altering the original meaning. No indicators of data exfiltration, malicious execution, or harmful prompt injection were found.
Capability Assessment
Purpose & Capability
The stated purpose, TTS-oriented Markdown/TXT cleanup with optional chapter splitting, quality checks, and reports, matches the SKILL.md instructions.
Instruction Scope
The instructions are task-focused and include safeguards such as preserving meaning, avoiding summarization, and making minimal changes when uncertain.
Install Mechanism
There is no install spec, no code files, no required binaries, no credentials, and the static scan reported no findings.
Credentials
The skill reads user-specified Markdown/TXT files or directories and writes cleaned output files and reports; this is purpose-aligned but users should scope paths and output locations carefully.
Persistence & Privilege
The artifacts do not request persistence, background execution, credentials, privileged configuration, or ongoing access.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install tts-cleanup-checklist
  3. After installation, invoke the skill by name or use /tts-cleanup-checklist
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of tts-cleanup-checklist: a unified tool for batch cleaning and optional chapter-splitting of Markdown/TXT texts for TTS reading. - Provides configurable cleaning levels for references, punctuation, numbers, English alias handling, and sentence length. - Ensures no content compression, no opinion injection, and minimal alteration of original meaning. - Stepwise cleanup from structure and noise removal to readability, digit oralization, and prosodic optimization. - Includes quality checks (reference residue, sentence length, chapter continuity, character count) and automatic generation of cleanup reports for both single and batch files.
Metadata
Slug tts-cleanup-checklist
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Tts Cleanup Checklist?

统一的 TTS 文本清洗与可选章节拆分技能。用于将 Markdown/TXT 清洗为可朗读版本,支持批量处理、可配置清洗档位、质检检查与报告输出。只要用户提到“清洗文本用于 TTS”“按清洗规则处理章节”“去参考文献/引用”“做可朗读稿”“批量清洗章节”,就应优先调用本技能。 It is an AI Agent Skill for Claude Code / OpenClaw, with 57 downloads so far.

How do I install Tts Cleanup Checklist?

Run "/install tts-cleanup-checklist" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Tts Cleanup Checklist free?

Yes, Tts Cleanup Checklist is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Tts Cleanup Checklist support?

Tts Cleanup Checklist is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Tts Cleanup Checklist?

It is built and maintained by booynal (@booynal); the current version is v1.0.0.

💬 Comments